Process development for flux-free soldering with preforms (Part 4)

In part four of the article series Process development for fluxless soldering we deal with the process window and process control.

Process window

Lastly, the process window needs to be defined. This step is necessary to determine the specification limits and action control limits. Variations in materials and soldering fixtures, but also aging of the soldering machine are factors that change the soldering profile over the course of time. Especially for statistical process control as well as data science, the upper and lower specification limits must be known to ensure a save and stable process. The following process steps are to be controlled:

  • Residual oxygen concentration during the entire process
  • Temperature distribution
  • Introduction time and concentration of the reduction agent
  • Peak-temperature und dwell time
  • Vacuum step time and final pressure
  • Cooling gradients and temperature homogeneity above and below liquidus

Design of experiments (DOE) can help to vary the parameters, followed by quality checks.

Process control and big data

The identified process limits are the basis to introduce statistical process control (SPC). In combination with control cards, the limits will be monitored. Depending on the product and the customer, data logger runs can be performed in intervals, ranging anywhere from daily to monthly.

The soldering process is not the only focus area when it comes to big data approaches. In fact, the complete manufacturing process and product life cycle are relevant; beginning with the supplier and ending with the finished product and how it is used by the customer. Big data can be used to solve short term production issues, but also to further improve and control a stable process in the long term (predictive analysis). A consistent traceability, respectively an appropriate data storage is crucial to successfully take advantage of big data. This includes an intelligent networking of different storage locations, such as the merging data from an enterprise resource planning system (ERP), a manufacturing execution system (MES) and a customer relationship management software (CRM). This can be achieved in a technically suitable timeframe by applying the CRISP-DM-Process diagram (CRoss-Industry Standard Process for data mining). The aim is a consistent data structure where the parameters are in columns and the equivalent results are in rows. Data mining from the data pool complements classical statistical methods. The more complex the data, the higher the likelihood that data mining tools are needed. Figure 2 explains this.

Abbildung 2: Figure 2: Difference between statistics and data mining, source: BNB Qualitätsstatistik und Training


The development of a paste-free soldering process with preforms and formic acid or hydrogen needs a highly systemic approach to achieve a stable process. When time and effort of the individual steps are reduced, from experience, this will need to be made up when dealing with customer complaints. The same also applies for data analysis and process control.

Do you have further questions about vacuum soldering? Then feel free to contact us.